Block unveils Dessa for direct audio emotion detection

Block, the company behind Square and Cash App, unveiled a new AI speech model called Dessa. According to The-decoder, the model was trained entirely on human speech without any text data. It processes audio directly to identify emotions, sarcasm, and pauses in real time. The model supports full-duplex communication and runs locally on a MacBook Pro M4 Max. Block is currently seeking AI security researchers to test the model for risks before it becomes publicly available.

Why it matters

Businesses in Switzerland can use this technology to build voice interfaces that react to customer sentiment. Most current systems rely on text conversion, which strips away tone. If your company handles high-touch customer service, a model that recognises sarcasm or frustration changes how you manage automated interactions.

• Automated support tools can detect escalation earlier by identifying vocal cues.
• Local processing on hardware like the MacBook Pro M4 Max allows for data privacy.
• Direct audio processing reduces the latency found in traditional speech-to-text pipelines.

What to watch

Monitor how other companies integrate these emotional cues into existing workflows. You should check if your current voice AI providers offer direct audio processing or if they still rely on text conversion. The ability to reason while maintaining a natural tone will determine if these tools provide actual utility or just a better user experience.

Businesses will deploy these interfaces by next year. This requires reliable real-time emotion detection and a commitment to local data processing for privacy.